Indirect genetic effects increase the heritable variation available to selection and are largest for behaviors: a meta-analysis

Abstract

The evolutionary potential of traits is governed by the amount of heritable variation available to selection. While this is typically quantified based on genetic variation in a focal individual for its own traits (direct genetic effects, DGEs), when social interactions occur, genetic variation in interacting partners can influence a focal individual's traits (indirect genetic effects, IGEs). Theory and studies on domesticated species have suggested IGEs can greatly impact evolutionary trajectories, but whether this is true more broadly remains unclear. Here, we perform a systematic review and meta-analysis to quantify the amount of trait variance explained by IGEs and the contribution of IGEs to predictions of adaptive potential. We identified 180 effect sizes from 47 studies across 21 species and found that, on average, IGEs of a single social partner account for a small but statistically significant amount of phenotypic variation (0.03). As IGEs affect the trait values of each interacting group member and due to a typically positive-although statistically nonsignificant-correlation with DGEs (r _DGE-IGE = 0.26), IGEs ultimately increase trait heritability substantially from 0.27 (narrow-sense heritability) to 0.45 (total heritable variance). This 66% average increase in heritability suggests IGEs can increase the amount of genetic variation available to selection. Furthermore, whilst showing considerable variation across studies, IGEs were most prominent for behaviors and, to a lesser extent, for reproduction and survival, in contrast to morphological, metabolic, physiological, and development traits. Our meta-analysis, therefore, shows that IGEs tend to enhance the evolutionary potential of traits, especially for those tightly related to interactions with other individuals, such as behavior and reproduction.

Keywords: animal model; associative genetic effects; interacting phenotypes; quantitative genetics; social evolution; social interactions.

Figure 1.

Indirect genetic effects occur when the genotype of one individual influences the expression of a phenotype in a conspecific, mediated by social interactions.

Figure 2.

Social heritability (social h²) from a phylogenetic multilevel meta-analysis with 40 studies and 146 effect sizes (Aim 1). Orchard plot showing the back-transformed meta-analytic mean as r, 95% CI (thick whisker), 95% prediction intervals (PI) (thin whisker), and individual effect sizes (r) scaled by their precision (semitransparent circles). k corresponds to the number of effect sizes, and the number of studies is shown in brackets. Note that uncertainty measures (i.e., 95% CI and 95% PI) are generated assuming a normal distribution around the meta-analytic mean, explaining why some intervals may overlap zero substantially despite the data analyzed being bounded between 0 and 1.

Figure 3.

Social heritability (social h²) meta-analytic estimates from phylogenetic multilevel uni-moderator meta-regressions testing the following moderators (Aim 2): (A) Trait category, (B) Age, (C) Sex, (D) Population type. Orchard plot showing the back-transformed meta-analytic mean as r, 95% CI (thick whisker), 95% PI (thin whisker), and individual effect sizes (r) scaled by their precision (semitransparent circles). k corresponds to the number of effect sizes, and the number of studies is shown in brackets. Note that uncertainty measures (i.e., 95% CI and 95% PI) are generated assuming a normal distribution around the meta-analytic mean, explaining why some intervals may overlap zero substantially despite the data analyzed being bounded between 0 and 1.

Figure 4.

IGEs and DGEs meta-analytic estimates from phylogenetic multilevel uni-moderator meta-regressions (Aim 3) comparing: (A) V_A vs. V_IGE, (B) narrow-sense h² vs. social h², and (C) I_DGE vs. I_IGE (evolvability). Orchard plot showing the back-transformed meta-analytic mean as r, 95% CI (thick whisker), 95% PI (thin whisker), and individual effect sizes (r) scaled by their precision (semitransparent circles). k corresponds to the number of effect sizes, and the number of studies is shown in brackets. Note that for (B) and (C), uncertainty measures (i.e., 95% CI and 95% PI) are generated assuming a normal distribution around the meta-analytic mean, explaining why some intervals may overlap zero substantially despite the data analyzed being bounded between 0 and 1.

Figure 5.

Meta-analytic estimates for Aim 4: (A) phylogenetic multilevel meta-regression comparing narrow-sense heritability (h²) to total heritable variance (τ²); (B) phylogenetic multilevel meta-analysis of r_DGE-IGE. Orchard plot showing the original (A) and back-transformed meta-analytic mean as r, 95% CI (thick whisker), 95% PI (thin whisker), and individual effect sizes (r) scaled by their precision (semitransparent circles). k corresponds to the number of effect sizes, and the number of studies is shown in brackets. Note that for (A) uncertainty measures (i.e., 95% CI and 95% PI) are generated assuming a normal distribution around the meta-analytic mean, explaining why some intervals may overlap zero substantially despite the data analyzed being bounded between 0 and 1.